Hacker News
points by NickNaraghi 1 day ago | 0 comments
Carbon is like Opus

Argon is like Sonnet

[–] qt31415926 1 day ago
I'm surprised no one at Google realized that both OpenAI and Anthropic's naming schemes follow a fairly clear progression and that they should do something similar to reduce confusion. I would have assumed the opposite given that Argon has a higher periodic number
[–] kylestanfield 1 day ago
Lots of commenters here are confused by Luna, Terra, and Sol still haha.
[–] Jtarii 1 day ago
It isn't obvious that a fable is "better" than an opus either.
[–] cubefox 1 day ago
Originally the model tier that was better than Opus was called Mythos, which makes sense. Fable makes no sense, since a fable is just a small story with animals. They should have used "Legend" rather than Fable.
[–] johanyc 1 day ago
That's on them imo. The point is consistency
[–] atakanaluc 1 day ago
Both seem to have the same pricing: $2 in / $10 out initially, rising to $4 in / $20 out after the introductory period. That made me wonder whether Carbon might be Argon with additional RL training. What’s the Opus/Sonnet comparison based on?
[–] owkman 1 day ago
No, it's not. The naming is just bad. It's the next iteration of Argon
[–] hiddencost 1 day ago
I don't think that's right. Carbon and Argon have the same price.

I suspect they're using a new naming convention because they had previous contractual obligations tied to the "pro" and "flash" naming, and by switching to a new naming schema they avoid those commitments.

[–] bsiverly 1 day ago
Helium will be small and squeaky I presume
[–] mattlondon 1 day ago
So Argon was 4.0 Flash effectively? Interesting that there is such a small gap between their new Pro and Flash tiers
[–] luckydata 1 day ago
Carbon is Fable / Astra level. Argon is their Opus / Sol.
[–] aslilac 1 day ago
then why are both priced like sonnet?
[–] fancyfredbot 1 day ago
They didn't pay the Nvidia tax
[–] luckydata 1 day ago
Carbon is discounted temporarily. It's twice as expensive as Argon. Google TPUs are more efficient than Nvidia hardware plus Google gets them at cost so they have both better margins and more pricing power compared to other frontier labs.